Diabetic foot osteomyelitis (DFO) is a leading cause of lower-extremity complications in individuals with diabetes, and timely, accurate screening is critical to prevent severe outcomes such as limb amputation. Although conventional radiography remains the most accessible imaging modality, the subtle and heterogeneous appearance of DFO often results in delayed or missed detection. Despite the rich morphological information encoded in foot radiographs, current deep learning methods tend to underperform in capturing localized pathological patterns due to architectural limitations. In this work, we propose Dual Backbone with Gated Fusion and Transformer encoder (DualBack-GFT), a deep learning framework for automated detection and localization of DFO in plain radiographs. The model leverages two complementary backbones, EfficientNet-B6 and ResNet-50, fused via a gated mechanism that adaptively combines image-specific features. These fused representations are further refined using transformer encoders, which effectively model long-range dependencies. The architecture operates in two stages: binary classification followed by confidence-weighted bounding-box localization. We evaluate DualBack-GFT on a curated, expert-annotated baseline dataset of diabetic foot X-rays with both diagnostic and bounding-box labels. The model achieves an AUC of 0.9683 and an average ground truth coverage of 62.71%, outperforming established baselines. These results underscore the potential of dual-stage, attention-enhanced models for interpretable and robust DFO assessment in clinical radiographs.

Training deep learning-based medical image segmentation models is challenging with limited curated datasets. For AGITG TOPGEAR, a gastric cancer trial, the Clinical Target Volume (CTV) is complex and defined by multiple anatomical landmarks, making upfront training data preparation difficult for an automated contour QA segmentation model. We investigate anatomical priors, derived from surrounding organ segmentations, to provide spatial context and improve TOPGEAR CTV segmentation accuracy. We also evaluate active learning, iteratively expanding the training dataset by selecting cases expected to improve performance. One hundred TOPGEAR CT scans were retrospectively analyzed. An initial set of 10 expert-contoured cases was used to train an nnU-Net model. TotalSegmentator generated a voxel-wise anatomical prior map from surrounding structures as an additional input channel. Active learning was simulated over four iterations, selecting cases by model uncertainty and segmentation performance. All models used five-fold cross-validation for an ensemble uncertainty measure. Evaluation used a hold-out testing set of 50 cases. The anatomical prior improved CTV segmentation accuracy, increasing mean Dice Similarity Coefficient (DSC) from 0.84 to 0.86. Active learning similarly improved performance to 0.86, with greatest benefit in the final round. Combining the anatomical prior with active learning achieved the highest accuracy, with a DSC of 0.87. Model uncertainty correlated with DSC, supporting its use in identifying suboptimal predictions and guiding active learning. Anatomical priors and active learning each improved CTV segmentation accuracy and generalizability, with their combination achieving the best performance, supporting integration into segmentation model development for automated contour QA in radiotherapy clinical trials.

BACKGROUND AND PURPOSE:Automated contour quality assurance (QA) has the potential to reduce resource and cost requirements for clinical trial QA. While several studies have developed such models, few have been translated for use in prospective real-world trials. This study aimed to implement a contour QA tool using a previously developed model and evaluate its performance during deployment within the TROG 18.01 NINJA prostate cancer trial. MATERIALS AND METHODS:A software tool was developed using an existing prostate clinical target volume (CTV) QA model and integrated into the trial QA workflow. A pilot study was conducted over an 18-month period, during which 56 CTVs were assessed. Reports generated by the tool flagged cases for review and were provided to radiation oncologists to support QA processes. RESULTS:All five protocol-violating CTVs were correctly identified during deployment, yielding a sensitivity of 1.0. However, a higher-than-expected false positive rate resulted in an accuracy of 0.46 and specificity of 0.41. Retrospective analysis showed that many cases submitted deviated from the model's training distribution, primarily due to inconsistencies in MRI acquisition and variation in submitted CTV definitions. Incorporating out-of-distribution detection based on histogram correlation and model uncertainty improved accuracy to 0.69 for in-distribution cases. Radiation oncologists reported time savings of up to 60 min per case. However, preparation of data remained time intensive for QA coordinators, highlighting the need for further workflow automation. CONCLUSION:These findings support the feasibility of automated contour QA in multicentre trials and offer guidance for future implementation at scale.

Multiparametric magnetic resonance imaging (mpMRI) is a key tool for assessing breast cancer progression. Although deep learning has been applied to automate tumor segmentation in breast MRI, the effect of sequence combinations in mpMRI remains under-investigated. This study explores the impact of different combinations of T2-weighted (T2w), dynamic contrast-enhanced MRI (DCE-MRI) and diffusion-weighted imaging (DWI) with apparent diffusion coefficient (ADC) map on breast tumor segmentation using nnU-Net. Evaluated on a multicenter mpMRI dataset, the nnU-Net model using DCE sequences achieved a Dice similarity coefficient (DSC) of 0.69 ± 0.18 for functional tumor volume (FTV) segmentation. For whole tumor mask (WTM) segmentation, adding the predicted FTV to DWI and ADC map improved the DSC from 0.57 ± 0.24 to 0.60 ± 0.21. Adding T2w did not yield significant improvement, which still requires further investigation under a more standardized imaging protocol. This study serves as a foundation for future work on predicting breast cancer treatment response using mpMRI.

Background and ObjectivesBio-medical image segmentation models typically attempt to predict one segmentation that resembles a ground-truth structure as closely as possible. However, as medical images are not perfect representations of anatomy, obtaining this ground truth is not possible. A surrogate commonly used is to have multiple expert observers define the same structure for a dataset. When multiple observers define the same structure on the same image there can be significant differences depending on the structure, image quality/modality and the region being defined. It is often desirable to estimate this type of aleatoric uncertainty in a segmentation model to help understand the region in which the true structure is likely to be positioned. Furthermore, obtaining these datasets is resource intensive so training such models using limited data may be required. With a small dataset size, differing patient anatomy is likely not well represented causing epistemic uncertainty which should also be estimated so it can be determined for which cases the model is effective or not.MethodsWe use a 3D Probabilistic UNet to train a model from which several segmentations can be sampled to estimate the range of uncertainty seen between multiple observers. To ensure that regions where observers disagree most are emphasised in model training, we expand the Generalised Evidence Lower Bound (ELBO) with a Constrained Optimisation (GECO) loss function with an additional contour loss term to give attention to this region. Ensemble and Monte-Carlo Dropout (MCDO) uncertainty quantification methods are used during inference to estimate model confidence on an unseen case. We apply our methodology to two radiotherapy clinical trial datasets, a gastric cancer trial (TOPGEAR, TROG 08.08) and a post-prostatectomy prostate cancer trial (RAVES, TROG 08.03). Each dataset contains only 10 cases each for model development to segment the clinical target volume (CTV) which was defined by multiple observers on each case. An additional 50 cases are available as a hold-out dataset for each trial which had only one observer define the CTV structure on each case. Up to 50 samples were generated using the probabilistic model for each case in the hold-out dataset. To assess performance, each manually defined structure was matched to the closest matching sampled segmentation based on commonly used metrics.ResultsThe TOPGEAR CTV model achieved a Dice Similarity Coefficient (DSC) and Surface DSC (sDSC) of 0.7 and 0.43 respectively with the RAVES model achieving 0.75 and 0.71 respectively. Segmentation quality across cases in the hold-out datasets was variable however both the ensemble and MCDO uncertainty estimation approaches were able to accurately estimate model confidence with a p-value < 0.001 for both TOPGEAR and RAVES when comparing the DSC using the Pearson correlation coefficient.ConclusionsWe demonstrated that training auto-segmentation models which can estimate aleatoric and epistemic uncertainty using limited datasets is possible. Having the model estimate prediction confidence is important to understand for which unseen cases a model is likely to be useful.

Cranial sex estimation often relies on visual assessments made by a forensic anthropologist following published standards. However, these methods are prone to human bias and may be less accurate when applied to populations other than those for which they were originally developed with. This study explores an automatic deep learning (DL) framework to enhance sex estimation accuracy and reduce bias. Utilising 200 cranial CT scans of Indonesian individuals, various DL network configurations were evaluated against a human observer. The most accurate DL network, which learned to estimate sex and cranial traits as an auxiliary task, achieved a classification accuracy of 97%, outperforming the human observer at 82%. Grad-CAM visualisations indicated that the DL model appears to focus on certain cranial traits, while also considering overall size and shape. This study demonstrates the potential of using DL to assist forensic anthropologists in providing more accurate and less biased estimations of skeletal sex.

Background and purpose: Previous studies on automatic delineation quality assurance (QA) have mostly focused on CT-based planning. As MRI-guided radiotherapy is increasingly utilized in prostate cancer treatment, there is a need for more research on MRI-specific automatic QA. This work proposes a clinical target volume (CTV) delineation QA framework based on deep learning (DL) for MRI-guided prostate radiotherapy.Materials and methods: The proposed workflow utilized a 3D dropblock ResUnet++ (DB-ResUnet++) to generate multiple segmentation predictions via Monte Carlo dropout which were used to compute an average delineation and area of uncertainty. A logistic regression (LR) classifier was employed to classify the manual delineation as pass or discrepancy based on the spatial association between the manual delineation and the network's outputs. This approach was evaluated on a multicentre MRI-only prostate radiotherapy dataset and compared with our previously published QA framework based on AN-AG Unet.Results: The proposed framework achieved an area under the receiver operating curve (AUROC) of 0.92, a true positive rate (TPR) of 0.92 and a false positive rate of 0.09 with an average processing time per delin-eation of 1.3 min. Compared with our previous work using AN-AG Unet, this method generated fewer false positive detections at the same TPR with a much faster processing speed.Conclusion: To the best of our knowledge, this is the first study to propose an automatic delineation QA tool using DL with uncertainty estimation for MRI-guided prostate radiotherapy, which can potentially be used for reviewing prostate CTV delineation in multicentre clinical trials.& COPY; 2023 Elsevier B.V. All rights reserved. Radiotherapy and Oncology 186 (2023) 1-6

Distal radius fractures (DRFs) are one of the most common types of wrist fracture and can be subdivided into intra- and extra-articular fractures. Compared with extra-articular DRFs which spare the joint surface, intra-articular DRFs extend to the articular surface and can be more difficult to treat. Identification of articular involvement can provide valuable information about the characteristics of fracture patterns. In this study, a two-stage ensemble deep learning framework was proposed to differentiate intra- and extra-articular DRFs automatically on posteroanterior (PA) view wrist X-rays. The framework firstly detects the distal radius region of interest (ROI) using an ensemble model of YOLOv5 networks, which imitates the clinicians' search pattern of zooming in on relevant regions to assess abnormalities. Secondly, an ensemble model of EfficientNet-B3 networks classifies the fractures in the detected ROIs into intra- and extra-articular. The framework achieved an area under the receiver operating characteristic curve of 0.82, an accuracy of 0.81, a true positive rate of 0.83 and a false positive rate of 0.27 (specificity of 0.73) for differentiating intra- from extra-articular DRFs. This study has demonstrated the potential in automatic DRF characterization using deep learning on clinically acquired wrist radiographs and can serve as a baseline for further research in incorporating multi-view information for fracture classification.

Tuberculosis (TB) is one of the leading causes of death worldwide. Automated detection of lung caverns associated with TB in Computed Tomography (CT) could help clinicians optimise treatment. However, caverns detection on 3D CT data is challenging due to the curse of dimensionality, thus requiring larger training data and more computational resource. Our team (AEHRC CSIRO) participated in ImageCLEFmed TB caverns detection 2022 to address this challenge, by developing a 2D YOLO-based model (TBdet-2D), and an efficient 3D Retina-U-Net-based model (TBdet-3D). Both networks were trained on 559 CT data with data augmentation and tested on 140 data provided by the challenge. The results show that TBdet-3D (mAP_IoU 0.504) outperformed TBdet-2D model (mAP_IoU 0.308) on testing data, indicating that employing a 3D approach instead of a 2D approach is more appropriate for the task. Our team placed first among the participating teams in this challenge. An overview of ImageCLEFmed Tuberculosis 2022 is available at: https://www.imageclef.org/2022/medical/tuberculosis.

Objectives: To conduct a systematic survey of published techniques for automated diagnosis and prognosis of COVID-19 diseases using medical imaging, assessing the validity of reported performance and investigating the proposed clinical use-case. To conduct a scoping review into the authors publishing such work. Methods: The Scopus database was queried and studies were screened for article type, and minimum source normalized impact per paper and citations, before manual relevance assessment and a bias assessment derived from a subset of the Checklist for Artificial Intelligence in Medical Imaging (CLAIM). The number of failures of the full CLAIM was adopted as a surrogate for risk-of-bias. Methodological and performance measurements were collected from each technique. Each study was assessed by one author. Comparisons were evaluated for significance with a two-sided independent t-test. Findings: Of 1002 studies identified, 390 remained after screening and 81 after relevance and bias exclusion. The ratio of exclusion for bias was 71%, indicative of a high level of bias in the field. The mean number of CLAIM failures per study was 8.3 ± 3.9 [1,17] (mean ± standard deviation [min,max]). 58% of methods performed diagnosis versus 31% prognosis. Of the diagnostic methods, 38% differentiated COVID-19 from healthy controls. For diagnostic techniques, area under the receiver operating curve (AUC) = 0.924 ± 0.074 [0.810,0.991] and accuracy = 91.7% ± 6.4 [79.0,99.0]. For prognostic techniques, AUC = 0.836 ± 0.126 [0.605,0.980] and accuracy = 78.4% ± 9.4 [62.5,98.0]. CLAIM failures did not correlate with performance, providing confidence that the highest results were not driven by biased papers. Deep learning techniques reported higher AUC (p < 0.05) and accuracy (p < 0.05), but no difference in CLAIM failures was identified. Interpretation: A majority of papers focus on the less clinically impactful diagnosis task, contrasted with prognosis, with a significant portion performing a clinically unnecessary task of differentiating COVID-19 from healthy. Authors should consider the clinical scenario in which their work would be deployed when developing techniques. Nevertheless, studies report superb performance in a potentially impactful application. Future work is warranted in translating techniques into clinical tools.

Research in artificial intelligence for radiology and radiotherapy has recently become increasingly reliant on the use of deep learning-based algorithms. While the performance of the models which these algorithms produce can significantly outperform more traditional machine learning methods, they do rely on larger datasets being available for training. To address this issue, data augmentation has become a popular method for increasing the size of a training dataset, particularly in fields where large datasets aren't typically available, which is often the case when working with medical images. Data augmentation aims to generate additional data which is used to train the model and has been shown to improve performance when validated on a separate unseen dataset. This approach has become commonplace so to help understand the types of data augmentation techniques used in state-of-the-art deep learning models, we conducted a systematic review of the literature where data augmentation was utilised on medical images (limited to CT and MRI) to train a deep learning model. Articles were categorised into basic, deformable, deep learning or other data augmentation techniques. As artificial intelligence models trained using augmented data make their way into the clinic, this review aims to give an insight to these techniques and confidence in the validity of the models produced.

Volume delineation quality assurance (QA) is particularly important in clinical trial settings where consistent protocol implementation is required, as outcomes will affect future as well current patients. Currently, where feasible, this is conducted manually, which is time consuming and resource intensive. Although previous studies mostly focused on automating delineation QA on CT, magnetic resonance imaging (MRI) is being increasingly used in radiotherapy treatment. In this work, we propose to perform automatic delineation QA on prostate MRI for both the clinical target volume (CTV) and organs-at-risk (OARs) by using delineations generated by 3D Unet variants as benchmarks for QA. These networks were trained on a small gold standard atlas set and applied on a multicentre radiotherapy clinical trial dataset to generate benchmark delineations. Then, a QA stage was designed to recommend 'pass', 'minor correction' and 'major correction' for each manual delineation in the trial set by thresholding its Dice similarity coefficient to the network generated delineation. Among all 3D Unet variants explored, the Unet with anatomical gates in an AtlasNet architecture performed the best in delineation QA, achieving an area under the receiver operating characteristics curve of 0.97, 0.92, 0.89 and 0.97 for identifying unacceptable (major correction) delineations with a sensitivity of 0.93, 0.73, 0.74 and 0.90 at a specificity of 0.93, 0.86, 0.86 and 0.95 for bladder, prostate CTV, rectum and gel spacer respectively. To the best of our knowledge, this is the first study to propose automated delineation QA for a multicentre radiotherapy clinical trial with treatment planning MRI. The methods proposed in this work can potentially improve the accuracy and consistency of CTV and OAR delineation in radiotherapy treatment planning.

Mammographic mass detection and segmentation are usually performed as serial and separate tasks, with segmentation often only performed on manually confirmed true positive detections in previous studies. We propose a fully-integrated computer-aided detection (CAD) system for simultaneous mammographic mass detection and segmentation without user intervention. The proposed CAD only consists of a pseudo-color image generation and a mass detection-segmentation stage based on Mask R-CNN. Grayscale mammograms are transformed into pseudo-color images based on multi-scale morphological sifting where mass-like patterns are enhanced to improve the performance of Mask R-CNN. Transfer learning with the Mask R-CNN is then adopted to simultaneously detect and segment masses on the pseudo-color images. Evaluated on the public dataset INbreast, the method outperforms the state-of-the-art methods by achieving an average true positive rate of 0.90 at 0.9 false positive per image and an average Dice similarity index of 0.88 for mass segmentation.

Previous studies on computer aided detection/diagnosis (CAD) in 4D breast magnetic resonance imaging (MRI) usually regard lesion detection, segmentation and characterization as separate tasks, and typically require users to manually select 2D MRI slices or regions of interest as the input. In this work, we present a breast MRI CAD system that can handle 4D multimodal breast MRI data, and integrate lesion detection, segmentation and characterization with no user intervention. The proposed CAD system consists of three major stages: region candidate generation, feature extraction and region candidate classification. Breast lesions are firstly extracted as region candidates using the novel 3D multiscale morphological sifting (MMS). The 3D MMS, which uses linear structuring elements to extract lesion-like patterns, can segment lesions from breast images accurately and efficiently. Analytical features are then extracted from all available 4D multimodal breast MRI sequences, including T1-, T2-weighted and DCE sequences, to represent the signal intensity, texture, morphological and enhancement kinetic characteristics of the region candidates. The region candidates are lastly classified as lesion or normal tissue by the random under-sampling boost (RUSboost), and as malignant or benign lesion by the random forest. Evaluated on a breast MRI dataset which contains a total of 117 cases with 141 biopsy-proven lesions (95 malignant and 46 benign lesions), the proposed system achieves a true positive rate (TPR) of 0.90 at 3.19 false positives per patient (FPP) for lesion detection and a TPR of 0.91 at a FPP of 2.95 for identifying malignant lesions without any user intervention. The average dice similarity index (DSI) is0.72±0.15for lesion segmentation. Compared with previously proposed lesion detection, detection-segmentation and detection-characterization systems evaluated on the same breast MRI dataset, the proposed CAD system achieves a favourable performance in breast lesion detection and characterization.

Breast mass detection and segmentation are challenging tasks due to the fact that breast masses vary in size and appearance. In this work, we present a simultaneous detection and segmentation scheme for mammographic lesions that is constructed in a sifting architecture. It utilizes a novel region candidate selection approach and cascaded learning techniques to achieve state-of-the-art results while handling a high class imbalance. The region candidates are generated by a novel multi-scalemorphological sifting (MMS) approach, where oriented linear structuring elements are used to sieve out the mass-like objects in mammograms including stellate patterns. This method can accurately segment masses of various shapes and sizes from the background tissue. To tackle the class imbalance problem, two different ensemble learning methods are utilized: a novel self-grown cascaded random forests (CasRFs) and the random under-sampling boost (RUSBoost). The CasRFs is designed to handle class imbalance adaptively using a probability-ranking based under-sampling approach, while RUSBoost uses a random under-sampling technique. This work is evaluated on two publicly available datasets: INbreast and DDSM BCRP. On INbreast, the proposed method achieves an average sensitivity of 0.90 with 0.9 false positives per image (FPI) using CasRFs and with 1.2 FPI using RUSBoost. On DDSM BCRP, the method yields a sensitivity of 0.81 with 3.1 FPI usingCasRFs and with 2.9 FPI using RUSboost. The performance of the proposed method compares favorably to the state-of-the-art methods on both datasets, especially on highly spiculated lesions.

This book constitutes the refereed joint proceedings of the Third International Workshop on Reconstruction and Analysis of Moving Body Organs, RAMBO 2018, the Fourth International Workshop on Breast Image Analysis, BIA 2018, and the First International Workshop on Thoracic Image Analysis, TIA 2018, held in conjunction with the 21st International Conference on Medical Imaging and Computer-Assisted Intervention, MICCAI 2018, in Granada, Spain, in September 2018. The 5 full papers (out of 10 submissions) presented at RAMBO, the 9 full papers (out of 18 submissions) presented at BIA, and the 20 full papers (out of 21 submissions) presented at TIA were carefully reviewed and selected. The RAMBO papers cover aspects of medical imaging where motion plays a role in the image formation or analysis. The BIA papers deal with topics such as computer-aided detection and diagnosis of breast cancer, quantitative analysis of breast imaging modalities, and large scale breast image screening and analysis. The TIA papers cover aspects of image analysis research for lung and cardiac diseases including segmentation, registration, quantification, modeling of the image acquisition process, visualization, validation, statistical modeling, biophysical lung modeling (computational anatomy), deep learning and novel applications.

Breast mass detection and segmentation are difficult tasks due to the variation in size and shape of breast masses. Constructing classifiers for this problem is also challenging due to the fact that normal tissue regions overwhelmingly outnumber abnormal regions. In this paper, we propose a novel approach for detecting and segmenting breast masses in mammography based on multi-scale morphological filtering and a self-adaptive cascade of random forests (CasRFs). CasRFs can cope with severe class imbalance by adding layers to the cascade until a minimum number of false-positives (FPs) is reached. The approach achieves an average sensitivity of 0.94 with 1.99 FPs/image on INbreast and a sensitivity of 0.77 with 3.93 FPs/image on DDSM BCRP.

PURPOSE:To develop a computer-aided detection (CAD) scheme for mass detection on digitized mammograms that achieves a high sensitivity while maintaining a low false positive (FP) rate using morphological enhancement and simple linear iterative clustering (SLIC) method.METHODS:The authors developed a multiple stage method for breast mass detection. The proposed CAD scheme consists of five major components: (1) preprocessing based on morphological enhancement, which enhances mass-like patterns while removing unrelated background clutters, (2) segmentation of mass candidates based on the SLIC method, which groups mass and background tissue into different regions, (3) prescreening of suspicious regions using rule-based classification that eliminates regions unlikely to represent masses, (4) potential lesion contour refinement based on distance regularized level set evolution, and (5) FP reduction based on feature extraction and an ensemble of undersampled support vector machines. Two datasets were built to design and evaluate the system: a mass dataset containing 187 cases (386 mammograms) and a nonmass dataset containing 88 mammograms. All cases were acquired from the digital database for screening mammography (DDSM). Approximately two thirds of the available masses were used for training the system, and the remaining masses and nonmass dataset were used for testing.RESULTS:Testing of the proposed CAD system on the mass dataset yielded a mass-based sensitivity of 98.55%, 97.10%, 92.75% at 0.84, 0.63, 0.55 FP mark/image, respectively. Tested on the nonmass dataset, the scheme showed a FP rate of 0.55, 0.34, 0.30 mark/image.CONCLUSIONS:The results indicate that the system is promising in improving the performance of current CAD systems by reducing FP rate while achieving relatively high sensitivity.

Breast cancer is one of the most common cancers among females worldwide. Early detection and diagnosis is the most effective way to reduce mortality. Mammography and breast magnetic resonance imaging (MRI) are two commonly used imaging modalities for breast cancer detection and diagnosis. Mammography is the primary screening tool for breast cancer detection. Breast MRI is often used as an adjunct to mammography for diagnostic purposes after suspicious findings are spotted on mammography. However, there are difficulties in the interpretation of both imaging modalities. Mammography is commonly used in population-based screening programs, and the large number of mammograms can be overwhelming to human readers given the shortage of radiologists in Australia. Breast MRI is difficult to interpret since it generates multiple anatomical and dynamic contrast enhanced (DCE) sequences from its 4D data acquisition, and it can be extremely challenging to visually analyze both enhancement kinetics and morphology of lesions at the same time. To assist radiologists in breast image interpretation, breast computer aided detection/diagnosis (CAD) systems have been developed to improve work efficiency and the confidence of decision.Breast lesion detection and segmentation are challenging since breast lesions vary in size, shape and contrast to the background. Previous studies mostly regard lesion detection and segmentation as separate tasks, only focusing on either detection or segmentation. Some others complete the workflow by stacking a segmentation stage onto a detection stage. However, the segmentation is often only performed on the true positive detections manually confirmed by the user, ignoring the false positive detections. It could be more helpful to the radiologists if the CAD systems have the ability to visualize the detection and segmentation results altogether, since shape and margin are important features for human readers to make detection and diagnostic decisions.This thesis presents a series of fully automatic mammographic CAD systems that integrate mass detection and segmentation, and a breast MRI CAD system that combines lesion detection, segmentation and characterization in one framework.The proposed mammographic CAD systems can detect and segment masses simultaneously without user intervention. Both the traditional workflow which involves unsupervised region candidate generation and hand-crafted features, and the deep learning (DL) based workflow are explored. A novel unsupervised region candidate generation technique, multi-scale morphological sifting (MMS), is proposed. The MMS uses morphological filters with oriented linear structuring elements (LSEs) to sieve out mass-like patterns including linear spicules that are normally present in breast masses. This technique can accurately extract masses as region candidates. Class imbalance is an issue that hinders the classification of masses, as the normal region candidates often outnumber the mass candidates. A novel cascaded random forests (CasRFs) and the use of random under-sampling boost (RUSboost) are proposed to tackle this problem. A novel DL-based framework that uses pseudo-color mammograms generated by the MMS and the Mask R-CNN is also proposed. A pseudo-color mammogram is composed with the original grayscale mammogram in one channel and two MMS enhanced images in the adjacent channels, which provides color contrast between masses and background tissue. The mammographic CAD based on the combination of pseudo-color mammograms and the Mask R-CNN achieves the best performance, yielding an average true positive rate (TPR) of 0.90 at 0.9 false positive per image (FPI) for mass detection and an average Dice similarity index (DSI) of 0.88 for mass segmentation. Evaluated on the same datasets as several previous studies, the mammographic CADs outperform the state-of-the-art CADs in both lesion detection and segmentation in a much simpler architecture.The proposed breast MRI CAD can simultaneously detect, segment and characterize lesions as malignant or benign. The MMS is extended into 3D MMS which can segment lesions as region candidates more accurately and efficiently than the brute force searching methods commonly used in previous studies on breast MRI images. Provided with accurately segmented lesion candidates, the system combines hand-crafted analytical features derived from all available image sequences, including the T1- and T2-weighted anatomical sequences and the DCE sequences, to achieve a comprehensive analysis of the morphology and contrast enhancement kinetics of the lesions. Hand-crafted features are adopted since it is still challenging for the DL-based methods to directly handle 4D multi-modal breast MRI data. Evaluated on the same dataset, the breast MRI CAD outperforms the state-of-the-art methods in lesion detection and identifies malignancies additionally, yielding a TPR of 0.90 at 3.19 false positives per patient (FPP) and a median DSI of 0.77 for lesion detection-segmentation, and a TPR of 0.91 at 2.9 FPP for malignancy identification.